Motion State Estimation Of Four-wheel Drive Electric Vehicles

With the development of information technology,more and more information technology methods have played an important role in the automotive industry.Since the 1990 s,researchers have devoted themselves to improving the performance,safety,and comfort of automobiles,and have developed more and more control systems.In the field of vehicle control,the estimation of the vehicle motion state is a very important part.The accuracy of the motion state estimation determines the quality of the control effect in the vehicle control system and driving assistance system.Kalman filter,Extend Kalman filter,Unscented Kalman filter and other filtering algorithms are widely used in the field of state estimation due to their low computational complexity,which can be used to deal with nonlinear problems and high computational accuracy.With the gradual development of neural networks,more and more researchers try to combine state estimation with neural networks to enrich the diversity of estimated parameters and improve the accuracy of estimation results,such as BP neural network,convolutional neural network,Recurrent neural network,etc.In the field of state estimation,the introduction of neural network can not only estimate the parameters of the vehicles’ motion state,but also quantify the perception of the vehicles’ environment such as road information.In this paper,we take the state estimation algorithm as the research object,and explores the application of the filtering-based state estimation algorithm and neural network in the field of state estimation.In order to improve the parameter accuracy of the state estimation algorithm,a state estimation algorithm based on filtering and tire stiffness correction is proposed;in order to enrich the parameter estimation diversity of the state estimation algorithm and quantify the road information,a neural-network based state estimation algorithm combined with classic algorithm is proposed,the specific work of this paper is as follows:1.The tire modeling methods and vehicle modeling methods commonly used in the field of state estimation are introduced.The commonly used filtering algorithms in the field of state estimation,such as classical Kalman filter,extended Kalman filter and other algorithms,and elaborated its mathematical principles,application scenarios and numerical iterative solution methods are introduced in detail.We introduced the basic principles of neural networks and the network structure of neural network algorithms such as BP neural network,convolutional neural network(CNN),recurrent neural network(RNN),LSTM,forward propagation and backward propagation,advantages and disadvantages,and state estimation Field application.In this paper,the filtering algorithm is improved and the neural network algorithm is applied to the field of state estimation.An improvement plan is proposed for the accuracy and diversity of parameter estimation,and it is pointed out that the introduction of neural network can greatly improve the effect of the classic state estimation algorithm.2.Model the tires and the entire vehicle,based on the extended Kalman filter and the "predict-correction" numerical calculation method,taking into account the inaccurate estimation due to changes in working conditions the correction coefficients along with CNN are introduced to improve the accuracy of parameter estimation.3.Design a recurrent neural network(RNN)and the corresponding input and output vectors for state estimation.Collect a large amount of data on different roads and under different working conditions for network training and parameter adjustment.Using LSTM neural network to improve the accuracy of parameter estimation and the overall robustness of the algorithm.4.Use real vehicles to collect data and train the network structure.The trained model was used for real vehicle testing,and a lot of different working conditions were designed for testing.The results show that the estimation errors are within a reasonable range and the robustness is good,and the estimation accuracy meets the requirements,which proves the effectiveness of the estimation algorithm proposed in this paper.